Thermo-electric Interpolation Formula. 



473 



Table II. 

 Barus's Air-Thermometer Comparisons, Series I.-V. 



t°. 



4* 



4e. 



Avenarius. 



Exponential. 



Logari 



thmic. 





Per cent. 





Percent. 





Per cent. 







s, 



§ 



o, 



8 



8, 



8 







Da.-Eq. 



= 100- 

 e 



Da.-Eq. 







= 100- 



Da,-Eq. 



=100- 



°C. 



mv. 



mv. 



mv. 





mv. 





mv. 





00 



(-150) 











00 



+23 



+13 







378-5 



3679 



3829 



-84 



-4-0 



-66 



-1-2 



-33 



-0-60 



4403 



4508 



4658 



4-18 



+0-28 



+18 



+0-28 



+30 



+0-47 



522-0 



5486 



5636 



-4 



-007 



-25 



-0-34 



-36 



-0 50 



58S-4 



6404 



6554 



+70 



+1-1 



+33 



+040 



+9 



+012 



6721 



7550 



7700 



4-110 



+ 1-5 



+60 



+0-64 



+26 



+0-28 



745-6 



8530 



8680 



+82 



+0-92 



+26 



+025 



-9 



-o-io 



840-1 



9898 



10048 



+ 101 



4-1-0 



+49 



+0-41 



+26 



+022 



946-6 



11396 



11546 



+9 



+ 0-07 



-19 



-0-14 



-13 



-0-10 



1019-7 



12475 



12625 



-45 



-032 



-45 



-0-32 



-10 



-0 07 



Average percenta 



ge deviati 



cms 



. 0-93 





053 





The Avenarius equation applied to these data yields 



So *= 9-104* + 3-249.HT 3 * 2 

 microvolts. 



Range 350° to 1075° C. 



Computing from this equation values of 2 £ for the 

 successive values of t in column 1 and subtracting them 

 from the data in column 3 gives the deviations between data 

 and equation. These are expressed in microvolts in column 

 4, and in percentages in column 5, the percentage being 

 reckoned in terms of e t as deduced by the exponential formula. 

 Objections may be felt to this use of e (here as throughout 

 the subsequent tables) as a basis, since e t involves e which is 

 an extrapolated value, certainly not exact, and possibly wide 

 of the truth. Such a criticism is valid, but inasmuch as the 

 values of e employed are nearly equal, and as the percentage 

 deviations are used merely for purpo.-es of expressing relative 

 accuracy, the possible error involved is nearly annulled. 

 Hence, although it would be better to compute 8 t, and express 

 this as a percentage of the absolute temperature t, the added 

 labour did not seem justified by the small gain, 



